Those aren't subscription related. The graph shows an increase in earnings from subscriptions despite lower overall numbers.
What does that mean? That many of the previous subscribers added to the total may have been 'one monthers' - people who only subbed for a single month in that quarter. Thus despite being one of the subscribers, they only contribute a third of revenue compared to a full-time player.
Example (not accurate at all, just to show a point):
1 million people pay for a single month's subs = $10,000,000
0.7 million pay pay for two months' subs = $14,000,000
This shows that the subscribers they have left in the western market are playing more consistantly during each quarter now. Its why the 'quarterly report' method is flawed if you only look at subscriber numbers and not overall revenue. 12 million subscribers only paying for a single month would bring in less revenue then 4.1 million paying for the full three months during a quarter, for example.